The most effective drinking-and-driving laws set prohibited BAC levels. However, most drivers have a very limited understanding of the BAC concept. Further, their ability to estimate their BAC levels is very limited. Evidence shows that over-the-limit drivers on the roads on weekend evenings underestimate their BAC levels. It appears, therefore, that providing heavy drinkers with a method of correctly estimating their BAC levels could reduce the risk such drivers pose to themselves and others. In the last 30 years, several aids for counting drinks or performing self-tests have been developed. Yet, few have been adequately evaluated and most have failed to become commercially viable in the United States. Recently, a new self-test device the Guardian Angel (GA)--has been developed that analyzes the alcohol content of saliva samples. The GA unit has many good features such as its size (small enough to be carried in the wallet) and cost (about $2 for four test strips). It has been advertised extensively and distributed through grocery stores and drug stores. Nonetheless, there is almost no information on the performance of the unit. The manufacturer reports that the device has been tested to the Department of Transportation standards, but only very limited information has been provided on their tests. Furthermore, little information is known about the "field usability" of the product. In a limited pilot test, PIRE found that the accuracy of that performance depends in part on the BAC of the user. The GA test requires proper administration and interpretation to produce accurate results. Thus, intoxication of the test user and dark lighting conditions may lead to inaccurate test results that could increase the risk to the user. The aim of the proposed research program is to conduct two studies that will determine the validity and reliability of the GA saliva test compared to a standard evidential breath-test device and to investigate the factors that will determine the extent to which the results of the test will affect user decisions about drinking and about drinking-and-driving. The research also will examine the impact of verified accurate versus inaccurate BAC feedback, as well as that of categorical versus precise BAC feedback on drinkers' perceptions of intoxication and impairment. This research not only has practical implications by validating the GA test, but also theoretical implications regarding the role of BAC feedback in general.